Comment 31 for bug 466575

Having done some tests with various drives, I have come to believe that it is actually safe to ignore the error messages produced by sg3-utils. I think so for the following reasons:

1. I could not find evidence that Hardy Heron (the release I used before Lucid) made use of sg3-utils to flush caches/stop devices before unmounting USB devices. And still, the filesystems were closed cleanly.
2. Even if Hardy made use of the problematic sg3-utils functions ("sg_ll_sync_cache_10" and "sg_ll_start_stop_unit"), they still would have failed, because the relevant code is identical.
3. Even though the sg3 functions fail, udisks-daemon still invokes the "umount" command, so the filesystem is not left in an inconsistent state, when the device is removed.

For these reasons, I slightly modified udisks-helper-drive-detach to simply not report back error messages to the user. It still invokes the error-producing sg3 functions, but does not pass the errors on to the user. This way, the functions are still called for devices where they can be executed successfully. And for the devices where they fail, it is not any different - only that the user does not know anything about this. Anyhow, I have come to believe that the errors are caused by a crappy implementation of the SCSI protocol by my USB disk, which is in arbitrary 3.5" disk in a cheap disk enclosure, which I bought separately. The controller of the enclosure seems to produce errors no matter what command you throw at it. :-/

I am attaching the modified version of "job-drive-detach.c", which only differs from the original version ("job-drive-detach.c.orig") by one statement (setting return value to success; line 213). For your convenience, I'm also attaching a precompiled version of "udisks-helper-drive-detach", which is responsible for passing the error to the user. If you are fed up with the error messages like me, place the file in /usr/lib/udisks and you won't be bothered anymore. Make sure to create a backup of the original file. Also you will have to place my modified version there, whenever an update overwrites it (which doesn't happen too often anyway).

Please note that this is not really a patch, as it does not fix the underlying problem. But I don't really mind, since it wasn't any different in Hardy. The only difference is that I didn't notice it back then and now I am bothered with it every time I unplug a device. If you think so, too, then you are welcome to use my modified version of "udisks-helper-drive-detach".